GPT Image 2 Prompt Updater 🔄

The prompt library in this pack isn't static - the upstream collection keeps growing, and every so often you'll want the new stuff. GPT Image 2 Prompt Updater is the pack's built-in update mechanism: run it and it rebuilds the local prompt database from two remote sources, downloads any new preview images, and refreshes every Selector and Preview node on your canvas. It's also the repair tool, because this is the node that fixes a library whose first-run data build went sideways.

This is the one node in the pack that actually leaves your machine - it reaches out over the network. Everything else is strictly local.

How it works

The node runs a two-stage sync, both done over plain HTTP (no git binary required, which matters on Windows installs where git isn't always available or reliable):

1. sync_github - invokes the pack's build_local_prompts.py script in a subprocess. That script downloads the latest READMEs from the upstream EvoLinkAI collection via GitHub raw, lists the image folders through the GitHub API, downloads only the images that changed (incremental), and recovers prompt text from git commit history when an upstream README has emptied its prompt blocks. Everything lands in data/.
2. sync_opennana - a second pass that pulls prompts from opennana.com using a sitemap-based incremental sync, tagging them as a separate source so the two don't clobber each other.

The output is a status STRING with the full report: how many prompts were added or removed before/after, which stages ran, and any errors. That's your changelog and your debugging log in one.

The seed input exists for a subtle reason: the node's IS_CHANGED returns NaN, the ComfyUI "always re-execute" signal, so it never gets skipped by caching - and the seed is a frontend convenience to give you a randomize control that forces the node to fire again on queue. Change it, hit run, done.

The inputs that matter

The current shipped inputs are sync_github (default on), sync_opennana (default on), and seed. The README in the repo still describes an older rebuild_from_readme / git_pull_first design - that's stale documentation; trust the node, which is what you'll actually see on the canvas. Leave both sync toggles on for a full refresh. If you only want your custom prompts updated or want to avoid a network call, you can't skip the remote sync per-node - but you don't have to run this node at all if the library is fine as-is.

Install and when you'll need it

Same pack as the rest - ComfyUI Manager search "GPT Image 2 Prompts", or git clone https://github.com/T8mars/comfyui-gpt-image2-prompt-T8 into custom_nodes/ and restart. No pip requirements.

You'll reach for the Updater in two situations. First, to grab new prompts: run it periodically with both toggles on. Second, to repair a broken library - if the Selector shows an empty dropdown, the first-run auto-build likely failed, and this node is the no-terminal fix. Two things to know going in: it needs working network access, and the first run can take a few minutes (the build script has a 600-second timeout). Behind a strict proxy or offline, it'll fail - the status string will tell you which stage died, and the pack's Execution Checker node can confirm the damage afterward.